The biggest bottlenecks to creating the theory of intelligence

Hey guys, I know the logic is straight forward that the best place to start building any intelligence is to study the only thing that is intelligent - our brain.

Why hasn’t there been an established theory on how the brain works already?

Is it the problem of not having enough experimental data? Is it the lack of better techniques to experiment?

Is it the lack of consensus that building a theory on intelligence is a tractable problem?

Many tech companies are now rushing into building hardware and software that are not really what we call intelligent. I think many of them are doomed to fail when we finally figure out the algorithm of the brain.

What is slowing down the current effort to building such a theory? What do you all think?

1 Like

I have my model and of a AGI brain. Been posting it on other forms for the young crowd
to learn. The theory and the model will stand on it own to the end of time. I was hopping
young crowd give me the respect like Jeff Hawking gets. But i i am finding out that the older
crowd copying it and not giving me any credit:(

So if there someone else has a complete model i do not see them coming forward with there
work. Unless they have a lot of personal money to finance the project on there own.

Finding the algorithms underlying the operation of the brain is tricky because of the distributed nature of the computations. That distributed nature arises because the structure of the brain can only be proscribed to low degree of precision by the information stored in DNA. It cannot say where every single brain cell should go and exactly what other brain cells it should be connected to. It can just say replicate a bunch of neurons here and there. An advantage is that provides a lot of redundancy in case for example you drink a bottle of md 2020 and a number of brain cells die as a result. Or worse yet consume a beverage or artificial sweetener containing methanol that will pickle your neurons in formaldehyde as it is converted.

A bottle neck? Intelligence and consciousness are two different thing.
Will a conscious machine need to come first before a intelligent machine.

Nobody here is stopping you. Code one up and hook it to a chatbot on the web or some other accessible demonstration model. If it works you will be a celebrity.

As far as JH - he has invested considerable funds and time to support his intuitions on how he thinks the brain works. He has been willing to change his mind as he moves forward and sees things from the perspective of having tested his theories. He has published and put his ideas in the wide world for people to shoot down. He has a paid staff to turn his ideas into a workable product. He travels the world promoting his work. Even with all that the acceptance of his efforts has not been quick or easy.

This is a very different level of involvement from what you are describing. Maybe there is more that you are not including in your description but from what I have read so far you have not put in the work to spread and defend your ideas. A few demonstrations and perhaps some code fragments are a very long way from a working AGI.

I am an engineer and inventor. As is the case with most engineers - my company requires that I innovate on command, over and over, as the need arises. What I have learned from this is those good ideas are a dime a dozen - pushing them through to a “finished” product takes much more than just having the idea.

Please keep working on your projects - we (the AI community) needs brilliant people sorting this all out. As you mentioned - nobody has all the answers at this point. You clearly have some great ideas. I do think you need to do more work to get to the AGI stage.


I think this is probably the bottleneck. In order to develop a working model of the brain (or cortex in this case) there has to be a good scientific consensus on the functional roles of cortical structures. This takes time because the techniques for probing the in-vivo brain is relatively limited.

However, it does seems as though there is a good theory emerging on how the cortex works - and its driven by neuroscience consensus. I do believe, however, that this theory is not near explaining intelligence. Human intelligence is due more to the prefrontal cortex, which is even more nebulous than the old cortices that neuroscience is gaining ground on.

1 Like

I liken this to trying to understand a television from poking at it with an oscilloscope. What part makes the pictures? Is there a face circuit? A sky circuit? How does it make the motion? Does it have a model of the world? If you do know how it works these suppositions seem silly and misguided. (NO NO NO - I am NOT proposing a soul that is channeled by the brain!) In the beginning, you might work out the transistors and get a good working theory of how they work. The linear ones vs. the switching one may be a confounding factor for a while. Are diodes transistor or not? FET transistors are different again. It may take a long time to work out that those big square transistors are in fact made of millions of little transistors. Working out that that is a complete computer in the middle of everything is another big chunk of understanding that could take a very long time. It would make things so much easier if you had a working channel of information to feed the inputs but you are studying the TV at this point and that may seem like a needless complication.

Brain research is a lot like that.

Until you get the right model your suppositions may capture parts of the picture but there will be odd parts that just don’t fit. Once you have the right model everything that comes to light will just seem to fit. Big parts of how the brain works are falling into place. The overall theme of maps, tracts, local processing, memory in synapses, flows of information over the various tract - these are well established and are certainly going to be part of whatever comes out of all this.

What is now under intense study is what information is being worked on within the brain. Various brain imaging techniques are telling volumes on the large-scale flows. Better probing is yielding previously unknown details like waves and grids on more local scales. None of this conflicts with prior assumptions - it just starts to fill in the details of what that local processing might be.

I think that there really is a little problem trying to get out and we just don’t know how to assemble the puzzle pieces. I liken this to chemistry before the Periodic table. Once they worked out how to arrange things mountains of previously known data fitted together into a harmonious whole. I suspect that both intelligence and consciousness will work this way - deep insight followed by consolidation of known facts.

I will add that the people that are shopping around some sub-subsystem of the brain as a solution to some problem are not necessarily going to go away after the “AGI solution” comes out. They have made a solution to problems using the bit of functionality that they have settled on. There is no reason to suspect that the problem they are solving will go away. In all likely hood, they will adopt the new AGI tools to enhance the solution to whatever problem space they are working in to make an even better product. They will have established a presence in that solution space and have the chance of remaining the best solution going forward.

1 Like

I’ve seen your links to posts on Google groups and Youtube. I have to say that I didn’t come across anything that explains your model in great depth, though (I could just be blind). What I’ve seen so far looks more like DNN computer vision than AGI, but maybe that is just due to a lack of more detailed information. Could also be a difference in definition…

I recommend you start your own thread here on the forum in the Community Lounge and go into your ideas in minute detail for the community here to review and comment on. Would also help to have more information describing what specific aspects of your model classify it as AGI.


The way I see it, better experiments would make the facts clearer, but the main issue is too little integration of results. There would be almost no disagreement if everyone automatically knew the results of every study. I’ve read almost no article which puts results together to produce clear facts about what the brain does. Some articles are part of an ongoing conversation about a very specific topic, but that isn’t enough.

For example, I’ve been researching POm (a thalamic nucleus) and how it fits into circuits for a while, but I’m still reading new things about it. It took two dozen articles before I read that it is actually two different nuclei (maybe three), which explains a lot of seemingly contradictory results. This sort of thing has happened a lot.

I don’t think we need to understand every fundamental aspect of the neocortex. But without summaries, even about low-level circuits, ideas about what the brain does take a long time to produce and test.


I think my favorite aggravation is that everyone has to call the same bits with different names.
Depending on who you are reading it’s “posteromedial complex (PoM)” or “rostral sector of the posterior group” or “posterior medial nucleus” or part of the “ventral nuclear group” which subdivides into VI, VPM, & VPL. You are just supposed to know all the names and the rather arbitrary groupings. And just to make things even more fun - the names change as you go from one animal to another.

1 Like

Even with a good start like yours there are honestly now so many (almost) AGI models that you will have to at least beat IBM Watson at Jeopardy or the young crowd will not be interested.

What a young scientist usually most needs is the simplest possible model to start from, fastest and easiest way to bring something to life in a virtual reality or a robot platform. What led me to the Numenta forum is the way HTM makes it easy to go from there with. Same overall interaction occurs, not a whole new model.

This weekend in a neuroscience forum I found a talented young student needing the reasons for their textbooks missing information required to make full sense of what remained. It took hours to get all the words right, but the time spent will at the same time explain the reasons for the way I apply the model to the genetic, cellular, as well as the multicellular “brain” level. In this case a serious classroom problem exists where teachers usually appreciate all help they can get from the outside in regards to topics concerning religious addiction, where parents may soon call to complain about being mentioned in a health class where at least one student needed to know. I normally do not need to go into this much detail and would rather not have to. In this case though a theory that still gives me eerie chills to have my name on is being used for what it’s intended for, I only need to show that:

The greatest bottleneck I know to a theory of intelligence for the whole human brain is the number of specialized circuits that each modify behavior but not fully control it. Even with such a theory the most very basics are still needed, a high school level starting point that right away has talented students on a rewarding science mission of a lifetime. That is something also able to stand “to the end of time” and in this case it very much influences religion, which slowly changes over time in a way that outlasts civilizations that regularly come and go with much of their scientific knowledge quickly lost with them. I try to think way way ahead, in part to avoid chaos from my not putting enough thought into unintended consequences of what can at first seem like trivial actions. What ends up lasting forever is often not was initially intended, so I try my best to not accidentally do that.

Yes i have a complete model of of human psychology that handles any issue with
mind and and body relationship. Ii is based on logic of the collective swarm.

The AGI will not be a chat bot or a Watson. Because it will be conscious. It will use my
definition of consciousness.
It will start out with the hardware that is close to a human child body. A personality will be grown withing the machines computer brain. It will use a starter code. That will grow. And it will also
need to learn from a human parent.
So this will take a long time to develop the personality within the machine. Years of work.
And may different implementation. Because most will have screw up mind that will
close resemble childhood psychological disorders. Just need to get one with the right
configuration and then it can be cloned in the millions.
I can guarantee you all that my twenty, plus, years on this is that elaborate, in theory, not
yet implemented in hardware or software.

My AGI brain will handle input and output into the real word like a human and will implement perfect human psychology. A perfect human mind. But the brain will not be human and will have few matching sections, regions, and parts.

Here is some trying to copy may work:

To me, it’s odd to say that others are copying you and not giving credit. It’s quite possible that other intelligent people such as yourself are arriving at similar conclusions (which is actually a good thing).

At the same time, ask what your motivation is. Is it that you become world famous and renowned? Or are you trying to serve some other higher goal of scientific discovery?

I agree with Mr. Lamb here… I would love to learn more about your take on all this, but you’re going to have to put work into it, starting a new thread and outline in minute (I’d say painstaking) detail how your perspective works, how different components interact, how information moves from one region to another (while not forgetting biological constraints, if you’re using a biological basis), then be open to constructive questions, criticisms, or challenges. Or maybe we’ll just love and accept it.

Make it easy to find your idea, in a single unified place, where you can keep updating and growing your post. We’d all benefit this way, including yourself.

Either way, that would be a more unified approach to consolidate your idea on AGI, rather than tidbits scattered about, or accusations of others copying your ideas (when it’s possible that they have come to the same conclusions independently).

For AGI, I’m personally open to any approach that works. Any AGI is going to be the culmination of giants standing on the shoulders of countless other giants, rather than a single-person act. Because of that, don’t shoot or disparage others who are working in the same ideas as you. United we stand and all that…


I would definitely like to see that theory elaboration. As for not yet implemented in software, is that something you are working on yourself, or you hoping to sell the theory to others with the required skills to implement it?

Sorry if focusing on this seems to be deviating from the original topic. I sort of feel like @keghn_feem’s case is a specific example of the original topic. From my perspective, the bottleneck in this case would appear to be communication of the theory in a detailed, cohesive format.

Sorry, I missed something. Can I have some details about that method? Many thanks.